Forecasting Chinese provincial CO2 emissions: A universal and robust new-information-based grey model

Abstract

Under China's new development philosophy, CO2 emissions forecasting is becoming more and more crucial for attaining carbon peaking and carbon neutrality targets. In light of this, this research suggests a new-information-based grey model built on cutting-edge methodologies that integrates the inventive damping accumulative generating operator, the data smoothing index, and particle swarm optimization. These three updates allow readers to anticipate the Chinese 30 provincial carbon emissions, which are characterized by insufficient information, regional heterogeneity, and complex patterns. For verification purposes, we investigate the robustness test referring to Monte-Carlo simulation and probability density analysis except as the multi-step-ahead forecasting. Extensive provincial experiments demonstrate that this proposed model considerably outperforms various prevailing competitors with remarkable universality, including grey models, artificial intelligence methods, and statistical models. Moreover, our critical empirical discovery is that this new model's performance fluctuates minimally across 30 provinces, with the average MAPE values less than 5% and 10% in the in-sample and out-of-sample periods, respectively, whereas the other benchmarks display unsteady simulation and prediction results. Furthermore, we employ the elite model to estimate Chinese 30 provincial CO2 emissions for the following three years, whose projections are endorsed by other studies and international organizations. Most importantly, our results provide insights for practitioners in formulating and implementing decarbonization policies and plans for specific provinces.